Devesh Singh

Deployed containerized AI services integrated into clinical FHIR/DICOM infrastructures, translating research machine learning prototypes into production pipelines in close collaboration with IT personnel and clinical stakeholders. Develop inference services for medical imaging models and ensure their reliable operation under real-world critical-infrastructure constraints.

DZNE is an interdisciplinary research institution that investigates neurodegenerative diseases. I developed end-to-end deep learning pipelines for large-scale MRI analysis, translating explainable AI research into clinical decision-support tools. Coordinated interdisciplinary collaboration between researchers, clinicians, and technical staff, supervised graduate students, and led projects from concept to deployment. Authored multiple peer-reviewed publications on applied AI.

Fellow of the Data Science for Social Good (DSSG) 2022 at TU Kaiserslautern, a full-time summer program training data scientists for social impact projects. In a team of 5, I worked with Paraguay's national procurement agency (DNCP) on detecting fraud in public procurement. My contributions included feature extraction, modeling, and pipeline building. Our team's solution improved fraud detection four-fold over DNCP’s current methods.

Research Questions: 1) How efficient are the image-to-image translation techniques which use deep generative networks, at bridging the gap between the synthetic and real domain datasets? 2) How effective is the overall synthetic data generation pipeline? Tasks: - Setting up and testing a pipeline for synthetic data generation - Using generative deep learning models: CycleGANs and CUT, for an image translation (computer vision) task Grade: 1.4

- Exploratory Data Analysis. - Modeling and forecasting inverter failure risks, using time series analysis. - Updating legacy Python repositories. - Code integration according to the Git-flow principles with Azure DevOps.

- Statistical testing of damage correlation to the usage. - Time series analysis of the machine signals. - Assisting the data collection and management process. - Analyze, visualize and infer knowledge from many different data sources.

Built explainable models, from a bio-medical data set to model tinnitus distress. Data Cleaning and Visualization, Exploratory Analysis, and Feature Selection were some of the steps involved in the process.

- Underwent a technical training in DotNet desktop and web application development. - Developed a holistic understanding of the financial system.

Grade: 1.3 Machine Learning | Deep Learning | Generative Modeling | Information Retrieval | Data Mining | Business Analytics | Database Management | Recommender Systems | Heuristic Optimization

Grade: 7.35/10 Software Engineering | Database Management | Computer Networks | OOPs (Java, C++) | Probabilistic Processes